Information prediction method and device, computer equipment, storage medium and program product

Abstract

The invention relates to the field of data processing technologies, text generation technologies, large model technologies and large language model technologies, and discloses an information prediction method and device, computer equipment, a storage medium and a program product. Obtaining interaction data of the first object identifier in the current session; acquiring historical task data executed by the second object identifier for the first object identifier, and determining to-be-confirmed data associated with the interaction data in the historical task data; and based on the data features of the interaction data, target task data associated with the consultation question in the to-be-confirmed data is predicted, and reply content corresponding to the consultation question is determined according to the target task data. According to the method, the target task data hitting the consultation problem in the historical task data can be predicted in combination with the interaction data, so that the prediction accuracy of the target task data is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical fields of data processing, text generation, large model, and large language model, and particularly relates to an information prediction method, apparatus, computer device, storage medium, and program product. Background Art

[0002] In an online platform for item interaction, a user can display item information of obtainable items that can be viewed by other users on the platform. Other users can browse the item information and obtain the required items based on the item information. Generally, the online platform also provides service functions for historical task data. After placing an order, the user can initiate a session through the service function to consult the user providing the item, or return or exchange the obtained item.

[0003] In the session scenario of the online platform, in order to improve the timeliness of content reply, the user providing the item usually uses an artificial intelligence assistant to perform a certain reply to the content consulted by the user. However, users often do not have the habit of actively sending the consulted task data to the consulted object. At this time, the artificial intelligence assistant often cannot understand the user's intention, so it cannot provide correct help to the user, thus affecting the user experience. Summary of the Invention

[0004] In view of this, the present disclosure provides an information prediction method, apparatus, computer device, storage medium, and program product to solve the problem that the artificial intelligence assistant cannot provide a correct reply to the user in the automatic reply scenario.

[0005] In a first aspect, the present disclosure provides an information prediction method, which includes:

[0006] After detecting a consultation question of a second object identifier for a first object, obtain the interaction data of the first object identifier in the current session, where the interaction data is used to indicate the session content and the session content source between the second object identifier and the first object;

[0007] Obtain the historical task data executed by the second object identifier for the first object identifier, and determine the data to be confirmed associated with the interaction data in the historical task data;

[0008] Based on the data characteristics of the interaction data, predict the target task data in the data to be confirmed that hits the consultation question, and determine the reply content corresponding to the consultation question according to the target task data.

[0009] In a second aspect, the present disclosure provides an information prediction apparatus, which includes:

[0010] An acquisition module, configured to acquire the interaction data of the first object identifier in the current session after detecting an inquiry question of the second object identifier for the first object, where the interaction data is used to indicate the conversation content and the source of the conversation content between the second object identifier and the first object;

[0011] A determination module, configured to acquire the historical task data executed by the second object identifier for the first object identifier, and determine the data to be confirmed associated with the interaction data from the historical task data;

[0012] A prediction module, based on the data characteristics of the interaction data, predicts the target task data in the data to be confirmed that hits the task data associated with the inquiry question, so as to determine the reply content corresponding to the inquiry question according to the target task data.

[0013] In a third aspect, the present disclosure provides a computer device, including: a memory and a processor, which are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to execute the information prediction method according to the first aspect or any corresponding embodiment thereof.

[0014] In a fourth aspect, the present disclosure provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to execute the information prediction method according to the first aspect or any corresponding embodiment thereof.

[0015] In a fifth aspect, the present invention provides a computer program product, including computer instructions, and the computer instructions are used to cause a computer to execute the information prediction method according to the first aspect or any corresponding embodiment thereof.

[0016] In the embodiments of the present disclosure, after detecting an inquiry question of the second object identifier for the first object, the interaction data of the first object identifier in the current session is acquired, where the interaction data is used to indicate the conversation content and the source of the conversation content between the second object identifier and the first object. Next, the historical task data executed by the second object identifier for the first object identifier can be acquired, and the data to be confirmed associated with the interaction data can be determined from the historical task data. Then, based on the data characteristics of the interaction data, the target task data in the data to be confirmed that hits the task data associated with the inquiry question can be predicted, so as to determine the reply content corresponding to the inquiry question according to the target task data, thereby predicting the target task data in the historical task data that hits the inquiry question by combining the interaction data, improving the accuracy of predicting the target task data, enabling the artificial intelligence assistant to provide correct help to the user, and further improving the user experience. Description of the Drawings

[0017] To more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the following will briefly introduce the accompanying drawings required for the description of the specific embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0018] Figure 1 is a flowchart of an information prediction method according to an embodiment of the present disclosure;

[0019] Figure 2 is a flowchart of predicting target task data according to an embodiment of the present disclosure;

[0020] Figure 3 is a flowchart of another information prediction method according to an embodiment of the present disclosure;

[0021] Figure 4 is a flowchart of predicting target task data based on a prediction model according to an embodiment of the present disclosure;

[0022] Figure 5 is a structural block diagram of an information prediction device according to an embodiment of the present disclosure;

[0023] Figure 6 is a schematic hardware structure diagram of a computer device according to an embodiment of the present disclosure. Specific Embodiments

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present disclosure with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are some, but not all, of the embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present disclosure.

[0025] It can be understood that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, usage scopes, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0026] For example, when responding to a user's active request, a prompt message is sent to the user to clearly prompt the user that the operation requested by the user will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, application program, server, or storage medium that performs the operations of the technical solutions of the present disclosure according to the prompt message.

[0027] As an optional but non-limiting implementation manner, in response to receiving an active request from a user, the manner of sending a prompt message to the user may be, for example, in the form of a pop-up window, and the prompt message may be presented in text in the pop-up window. In addition, the pop-up window may also carry selection controls for the user to choose to "agree" or "disagree" to provide personal information to the electronic device.

[0028] It can be understood that the above notification and the process of obtaining user authorization are only illustrative and do not limit the implementation manner of the present disclosure. Other manners that comply with relevant laws and regulations can also be applied to the implementation manner of the present disclosure.

[0029] It can be understood that the data involved in the technical solution (including but not limited to the data itself, the acquisition or use of the data) should comply with the requirements of the corresponding laws, regulations and related provisions.

[0030] Combined with the application scenario on which the execution of the information prediction method depends, the application scenario will be described herein.

[0031] In an online platform for item interaction, a user can display item information of obtainable items that can be viewed by other users on the platform. Other users can browse the item information and obtain the required items based on the item information. Generally, the online platform also provides service functions for historical task data. After placing an order, the user can initiate a session through the service function to consult the user who provides the item, or return or exchange the obtained item.

[0032] In the session scenario of the online platform, in order to improve the timeliness of content reply, the user who provides the item usually uses an artificial intelligence assistant to perform a certain recovery on the content consulted by the user. However, users often do not have the habit of actively sending the consulted task data to the consulted object. At this time, the artificial intelligence assistant often cannot understand the user's intention, so it cannot provide correct help to the user, thus affecting the user experience.

[0033] For example, when the artificial intelligence assistant tries to understand the user's intention during the session, it can perform a certain degree of prediction on the user's consulted task data. Here, a primary traditional tree model, such as a Gradient Boosting Decision Tree (hereinafter simply referred to as GBDT), can be set to be used for predicting the focus task data in the session. Specifically, the historical task data of the user at the order placement object can be obtained, and according to the task data status, creation time, update time, etc. of each historical task data, the probability of each historical task data being the focus task data in the current session can be predicted, and probability sorting is performed among the historical task data, and the historical task data with the highest probability is used as the focus task data.

[0034] However, this task data prediction scheme does not consider the actual interaction content between the user and the order placement object, such as the conversation content, etc., lacks horizontal comparison between different historical task data. When the differences between historical task data are relatively subtle, if only relying on information such as the order placement time of task data to predict the focus task data, the probabilities of the two task data being predicted as the focus task data may be very close or even the same, and it is impossible to accurately determine which task data the user actually cares about. For example, if the task data obtained by the user at the same time includes black pants and gray pants, and the user asks "Why did I only receive the black ones?". Here, through comparative analysis, it can be found that what the user actually wants to ask is "Why hasn't the gray pants arrived yet". Therefore, the focus task data should be the task data of "gray pants", but because the order placement time and other information of the two task data are similar, and the probabilities output by GBDT are similar, it is impossible to accurately determine the focus task data that the user wants to consult.

[0035] Based on this, the embodiments of the present disclosure provide an information prediction method. After detecting an inquiry question of the second object identifier regarding the first object, the interaction data of the first object identifier in the current session is obtained, where the interaction data is used to indicate the conversation content and the source of the conversation content between the second object identifier and the first object. Next, the historical task data executed by the second object identifier for the first object identifier can be obtained, and the data to be confirmed associated with the interaction data is determined from the historical task data. Then, based on the data characteristics of the interaction data, the target task data associated with the inquiry question is predicted from the data to be confirmed, so as to determine the reply content corresponding to the inquiry question according to the target task data, thereby predicting the target task data that hits the inquiry question in the historical task data in combination with the interaction data, improving the accuracy of target task data prediction, enabling the artificial intelligence assistant to provide correct help to the user, and further improving the user experience.

[0036] According to the embodiments of the present disclosure, an embodiment of an information prediction method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0037] According to the embodiments of the present disclosure, an embodiment of an information prediction method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0038] In this embodiment, an information prediction method is provided, which can be used in a computer device. Figure 1 It is a flowchart of the information prediction method according to an embodiment of the present disclosure. As Figure 1 shown, the process includes the following steps:

[0039] Step S101, after detecting an inquiry question of the second object identifier for the first object, obtain the interaction data of the first object identifier in the current session, where the interaction data is used to indicate the conversation content and the source of the conversation content between the second object identifier and the first object.

[0040] In an embodiment of the present disclosure, the second object and the first object can perform item interaction and conversation through a target platform. Among them, the target platform can be an online platform for item interaction. The first object can be a user who provides items that can be obtained in the target platform, and the second object can be a user in the target platform who has placed an order for an item with the first object or is about to place an order for an item. The second object identifier is the identity identifier of the second object in the target platform.

[0041] Here, the second object identifier can trigger the session identifier set by the first object in the target platform to open a dialog box with the first object. Here, the session identifier can be a public permission, that is, whether the second object who has placed an order with the first object can trigger the session identifier to jump to the conversation interface and have a conversation with the first object through the dialog box.

[0042] After detecting an inquiry question initiated by the second object identifier to the first object in the target platform, obtain the interaction data of the interaction operation between the first object and the second object identifier. For example, the interaction operation can include session behavior, order placement behavior, browsing behavior of item information displayed for the first object, etc.

[0043] Step S102, obtain the historical task data executed by the second object identifier for the first object identifier, and determine the data to be confirmed associated with the interaction data from the historical task data.

[0044] In an embodiment of the present disclosure, the historical task can be the data of the historical task of the second object identifier obtaining items from the first object within a historical period. Here, the historical task data can be preliminarily screened through the interaction data of the interaction operation to obtain the data to be confirmed, thereby reducing the computational complexity of predicting the target task data and reducing the demand for computing resources.

[0045] For example, when the above session behavior included in the interaction operation, the session content of the session behavior can be obtained as the interaction data, and based on the historical task data mentioned in the session content, the historical task data can be used as the data to be confirmed.

[0046] Step S103: Based on the data characteristics of the interaction data, predict the target task data in the data to be confirmed that hits the target task associated with the consultation question, so as to determine the response content corresponding to the consultation question according to the target task data.

[0047] In the embodiments of the present disclosure, the target task data in the data to be confirmed that hits the consultation question can be predicted based on a prediction model. Here, the prediction model may include a language model, and the language model can predict the probability of each data to be confirmed as the target task data based on the interaction data.

[0048] Specifically, the prediction model can first preprocess the interaction data, such as removing noise, outliers, and incomplete data, to improve the data quality. Then, feature extraction can be performed on the interaction data, that is, useful features are extracted from the interaction data, and the data is converted into a form suitable for model processing. For example, when the interaction data includes session text, features such as word vectors and syntactic structures can be extracted, and these features can better represent the essence of the data and provide a basis for subsequent matching calculations.

[0049] Next, the language model can extract the request features corresponding to the consultation question and calculate the similarity between the request features and the data features of the behavior data. For example, the cosine similarity algorithm can be used to calculate the similarity between the features. The cosine similarity algorithm measures the similarity by calculating the cosine value of the included angle between two feature vectors and outputs the target task data whose similarity meets the output conditions.

[0050] After determining the target task data, the task information corresponding to the target task can be sent to the second object identifier as the session content. Specifically, the task information may include the task card, task status, task logistics information, etc. of the target task data.

[0051] In addition, the consultation question of the second object identifier can also be replied based on the target task. For example, if the consultation question of the second object identifier is the logistics stage of the target task, then the task logistics information can be queried based on the target task data of the target task to obtain the logistics stage and reply to the second object identifier.

[0052] As can be seen from the above description, in the embodiments of the present disclosure, after detecting an inquiry question of the second object identifier for the first object, the interaction data of the first object identifier in the current session is obtained, where the interaction data is used to indicate the conversation content and the source of the conversation content between the second object identifier and the first object. Next, the historical task data executed by the second object identifier for the first object identifier can be obtained, and the data to be confirmed associated with the interaction data can be determined from the historical task data. Then, based on the data characteristics of the interaction data, the target task data associated with the inquiry question can be predicted in the data to be confirmed, so as to determine the reply content corresponding to the inquiry question according to the target task data, thereby predicting the target task data that hits the inquiry question in the historical task data in combination with the interaction data, improving the accuracy of predicting the target task data, enabling the artificial intelligence assistant to provide correct help for the user, and further improving the user experience.

[0053] In some alternative embodiments, the interaction operations corresponding to the interaction data include: the session trigger behavior of the second object identifier and the conversation behavior between the second object identifier and the first object, where the session trigger behavior is used to indicate the behavior path for the second object identifier to trigger the session. The above step S101 of obtaining the interaction data of the first object identifier in the current session includes:

[0054] Step S11, determining whether it is detected that the conversation behavior includes the sending behavior of the information entity; if so, step S12 is executed, and if not, step S13 is executed.

[0055] Step S12, obtaining the first historical task data and / or the first item information sent by the second object identifier to the first object during the conversation.

[0056] Step S13, obtaining the behavior path data corresponding to the session trigger behavior, where the behavior path data is included in the interaction data.

[0057] In the embodiments of the present disclosure, the information entity can be a card for displaying information. For example, a task card or an item card. The task card may include item information such as the item id of the item ordered by the user, the task id, the object id of the first object providing the task, the item name, etc., order information such as the order time, note information, task status, and logistics information. The item card may include picture information of the item, title information such as the item name, item value information, item description information, user evaluation information, etc.

[0058] Here, as Figure 2The figure shows a schematic flow diagram for predicting target task data. Among them, first, it can be detected whether there is historical task data of the user at the current first object. If not, it means that the second object identifier has not placed an order at the first object, and there is no need to perform the prediction of the target task data. If so, the historical task data is filtered according to the behavior data.

[0059] Specifically, when filtering the historical task data according to the behavior data, first, it can be determined whether the second object identifier has sent an information entity to the first object. If so, the information entity is obtained. For example, if the information entity includes a task card, the first historical task data can be determined based on the task card. If the information entity includes an item card, the first item information can be determined based on the item card.

[0060] In addition, when it is determined that the second object identifier has not sent an information entity to the first object, the behavior path data corresponding to the session trigger behavior can be obtained. Among them, the behavior data can include the incoming line entrance when the second object identifier triggered the session last time. For example, the incoming line entrance can be the session trigger identifier in the item card or the session trigger identifier in the historical task card.

[0061] Specifically, when the interaction data includes the behavior path data, the above step S102 of determining the data to be confirmed associated with the interaction data in the historical task data includes the following process:

[0062] Step S21, based on the behavior path data, determine whether the second object identifier triggers a session behavior through the historical task data. If so, execute step S22. If not, execute step S23.

[0063] Step S22, determine the historical task data triggered by the second object identifier as the data to be confirmed.

[0064] Step S23, when it is determined based on the behavior path data that the second object identifier triggers a session behavior through the item display entity, determine the data to be confirmed that matches the item display entity in the historical task data, where the item display entity is used to display the item information of the available items.

[0065] In the embodiment of the present disclosure, if the incoming line entrance indicated by the behavior data includes the session trigger identifier in the historical task card, the historical task data corresponding to the historical task card on this behavior path can be determined as the data to be confirmed.

[0066] For example, if it is determined from the behavioral data that the behavioral trajectory of the second object identifier is: opening task card 1 of historical task 1 in the historical task list, and jumping to task card 2 of historical task 2 through this task card 1, and finally initiating a consultation question through the session trigger identifier of task card 2, then the task data of historical task 1 and historical task 2 can be determined as data to be confirmed.

[0067] If the incoming line entry indicated in the behavioral data does not include the session trigger identifier in the historical task data card, it can be determined whether the incoming line entry indicated by the behavioral data includes the session trigger identifier in the item card (that is, whether the second object identifier triggers a session behavior through the item display entity, where the item display entity can be an item card). If so, the item id corresponding to the item card on this behavior path can be obtained, and the data to be confirmed that matches this item id can be determined in the historical task data.

[0068] For example, if it is determined from the behavioral data that the behavioral trajectory of the second object identifier is: opening item card 1 displayed by the first object, and jumping to item card 2 displayed by the first object through this item card 1, and finally initiating a consultation question through the session trigger identifier of item card 2. Here, if the item id corresponding to historical task 3 is the same as the item id included in item card 2, then historical task 3 can be determined as the data to be confirmed.

[0069] In addition, if the incoming line entry indicated by the behavioral data does not include the session trigger identifier in the item card, that is, the second object identifier has not sent an information entity to the first object and has not initiated a consultation question through the task data card or the item card, then the task data in the historical task data that meets the preset time condition is determined as the data to be confirmed.

[0070] Specifically, a time condition can be set in advance to screen the historical task data to obtain the data to be confirmed, thereby reducing the amount of computation when predicting the target task data based on the data to be confirmed. For example, if the historical task data includes all the task data of the second object identifier placing an order with the first object, and the preset time condition is 180 days, then the historical task data of placing an order within 180 days can be determined as the data to be confirmed that meets the preset time condition.

[0071] In the embodiments of the present disclosure, the interaction operation may include the session trigger behavior of the second object identifier and the session behavior between the second object identifier and the first object, so as to screen the historical task data based on the path data and session content of the behavior path, thereby reducing the amount of computation required to predict the target task data, reducing the dependence on computing resources, and further reducing the platform operation cost.

[0072] In this embodiment, another information prediction method is provided, which can be used in a computer device.Figure 3 is a flowchart of an information prediction method according to an embodiment of the present disclosure, such as Figure 3 As shown, the process includes the following steps:

[0073] Step S301, after detecting the consulting question of the second object identifier for the first object, obtain the interaction data of the first object identifier in the current session, wherein the interaction data is used to indicate the content of the session between the second object identifier and the first object and the source of the content of the session. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0074] Step S302: Obtain historical task data executed by the second object identifier for the first object identifier, and determine the to-be-confirmed data associated with the interaction data in the historical task data. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0075] Step S303, based on the data features of the interaction data, predict the target task data associated with the consulting question in the data to be confirmed, so as to determine the reply content corresponding to the consulting question according to the target task data.

[0076] Specifically, the interaction data includes: historical conversation content and interaction task information, wherein the historical conversation content includes at least one round of conversation content between the second object identifier and the first object, and the interaction task information is used to indicate the task information of the historical task data of the second object identifier initiating a service request to the first object. The above step S303 includes:

[0077] Step S3031, extracting data features of historical conversation content and interactive task information.

[0078] Step S3032, input the data features into the prediction model to obtain the target task data.

[0079] In the disclosed embodiment, the interactive task information may be the task information of the historical task data in which the second object identifier in the historical task data has initiated a service request such as a return. In addition, the conversation content initiated by the current consultation question may be determined as the current conversation content, and the conversation content of the previous round of the current conversation in the dialog box may be determined as the historical conversation content.

[0080] For example, if the entire conversation content between the second object identifier and the first object includes:

[0081] Session start time: 2024-07-15 12:05:58.

[0082] Second object identification: [Send item card] (item id: P012266).

[0083] First Object: It's my pleasure to serve you. What can I do for you?

[0084] First Object: Honey, please describe your problem with this product. I'll help you handle it.

[0085] Second Object Identification: Can it be changed to black?

[0086] First Object: Hello, honey. Usually, the item attributes cannot be modified. You can cancel the task data or apply for after-sales service according to the task data status.

[0087] First Object: You can place a refund order when placing the order.

[0088] Second Object Identification: But there is no black option in this.

[0089] First Object: Darling, let me help you check the problem, okay? Please wait a moment. Thank you!

[0090] Second Object Identification: [Send a picture] (This picture is the item parameter selection interface).

[0091] First Object: Dear, I've received the picture you sent. Can you describe your problem in words first? For example: The express delivery is too slow, the package is short of items, the item is damaged, or there is a quality problem with the item.

[0092] First Object: Hello, honey. Usually, the product attributes cannot be modified. You can cancel the task data or apply for after-sales service according to the task data status.

[0093] Second Object Identification: Then I'll apply for a refund now and change the color.

[0094] First Object: Mhm.

[0095] Conversation End Time: 2024-07-15 12:12:33.

[0096] Conversation Start Time: 2024-07-15 12:22:27.

[0097] Second Object Identification: [Send an item card] (Item ID: P012266).

[0098] Second Object Identification: Is this okay?

[0099] Second Object Identification: [Send a picture] (This picture is the item parameter selection interface).

[0100] Second Object Identification: Is it like this?

[0101] First Object: I'm here. Wait a moment while I take a look.

[0102] Session end time: 2024-07-15 12:24:10.

[0103] Here, the session between 12:05:58 and 12:12:33 can be the historical session content, and the session between 12:12:33 and 12:24:10 can be the current session content initiated again by the second object identifier after exiting the session interface by using the above consultation questions.

[0104] As Figure 2 can be seen, after determining the data to be confirmed, it can be determined whether the number of data to be confirmed is greater than 1. If not, it is determined whether the number of the filtered data to be confirmed is equal to 1. If so, the data to be confirmed is determined as the target task data. If not, it is determined that there is no target task data. Additionally, when it is determined that the number of data to be confirmed is greater than 1, the above prediction model can be called to predict the target task data.

[0105] Specifically, the above prediction model includes a recommendation model and a language model. The above step S3032 of inputting the data features into the prediction model to obtain the target task data includes:

[0106] Step a1: Input the data features into the recommendation model to obtain the prediction results corresponding to each data to be confirmed.

[0107] Step a2: When it is determined that the prediction result meets the adjustment condition, call the language model for the historical output result of the data to be confirmed, where the historical output result is used to indicate the prediction result of the historical task data discussed for the historical session content output by the language model.

[0108] Step a3: Adjust the prediction result based on the historical output result, and determine the target task data whose prediction result meets the matching degree condition among the data to be confirmed.

[0109] In the embodiment of the present disclosure, the recommendation model can be a BertFM (Bert + FM, where Bert is Bidirectional Encoder Representations from Transformers, meaning bidirectional encoder representations converter, and FM is Factorization Machine, meaning factorization machine) model.

[0110] Here, as Figure 4 shown is a schematic flowchart of predicting the target task data based on the prediction model. Among them, the interaction data can include, in addition to the above historical session content and interaction task information, the task information of the data to be confirmed.

[0111] After obtaining the interaction data, the data features of the interaction data can be extracted. The data features can include text features and FM features. Among them, the text features can be the text features of the historical conversation content, and the FM features can be the features of the interaction task information and the task information of the data to be confirmed, such as task status, task start time, and other features. Here, the specific method of extracting data features is subject to what can be achieved, and the present disclosure does not limit this.

[0112] Next, the data features can be output to the above-mentioned recommendation model so that the recommendation model outputs the scoring results corresponding to each data to be confirmed, that is, the above-mentioned prediction results. Here, when the differences in the prediction results are all within the preset prediction interval, it is determined that the prediction results meet the adjustment conditions. For example, if the scoring results are all within the preset score range, it is considered that the prediction results meet the adjustment conditions, that is, the differences between the data to be confirmed are small and further judgment is needed. For example, if the preset score range is [0, 0.5] and the scoring results include: 0.3, 0.5, 0.1, it is considered that the prediction results meet the adjustment conditions.

[0113] Here, considering that the language model requires a large amount of computing resources, if the language model is called in real time to adjust the prediction results, it will take a long time and cannot reply to the second object identifier in time. Therefore, the language model can be called asynchronously.

[0114] Specifically, considering that the focus of the dialogue consultation of the second object identifier in the same dialogue interface usually does not change, therefore, the language model can be called asynchronously for the prediction results of the focus task data discussed in the historical conversation content, that is, the above-mentioned historical output results. Here, if the historical output results are obtained as empty, the prediction results can be not adjusted. If not empty, the historical task data matching the focus task data in the historical output results is determined among the data to be confirmed, and the scoring result of the historical task data is adjusted.

[0115] After adjusting the prediction results based on the historical output results, the data to be confirmed whose prediction results meet the matching degree conditions can be determined as the target task data. For example, the matching degree conditions can include a score threshold. Here, the data to be confirmed whose scores in the prediction results exceed the score threshold can be determined as the target task data. Another example is that the data to be confirmed with the highest score in the prediction results can be determined as the target task data.

[0116] In the embodiments of the present disclosure, considering that the traditional Bert model has relatively strict restrictions on the input length of the text (the maximum input text length is 512), truncation processing is required in the case of longer conversation texts. However, in the present disclosure, the input upper limit is greatly extended through the language model, which can meet the prediction requirements of the task data in the present disclosure. In addition, considering that the semantic understanding ability and reasoning ability of the BertFM model are limited, resulting in poor accuracy of the prediction results, the prediction results can be adjusted through the language model to increase the confidence of the prediction results. At the same time, since the language model has high requirements for the computing resources of the device and the prediction takes a long time, which cannot meet the latency requirements of online real-time prediction. Therefore, in the present disclosure, the historical output results of the language model can be called asynchronously to reduce the prediction latency and at the same time reduce the requirements for computing resources.

[0117] In some alternative embodiments, step a3 above, adjusting the prediction result based on the historical output result, includes:

[0118] Increase the preset matching value for the prediction result corresponding to the data to be confirmed that matches the historical output result, and decrease the preset matching value for the prediction result corresponding to the data to be confirmed that does not match the historical output result.

[0119] Specifically, the preset matching value can be determined according to the scoring range of the prediction result, and the preset matching value can be greater than or equal to the scoring range. For example, if the scoring range is 0-1, the preset matching value can be 1, or a value greater than 1.

[0120] For example, if the focus task data included in the historical output result is historical task data 1, and the data to be confirmed includes historical task data 1, historical task data 2, and historical task data 3, where the prediction result corresponding to historical task data 1 can be 0.4, the prediction result corresponding to historical task data 1 can be 0.5, and the prediction result corresponding to historical task data 1 can be 0.1. Then, the prediction result of historical task data 1 can be increased by 1, and the prediction results of historical task data 2 and historical task data 3 can be decreased by 1.

[0121] In the embodiments of the present disclosure, the prediction results of the recommendation model can be adjusted based on the language model to make full use of the long text reasoning ability of the language model and improve the confidence of the finally predicted target task data.

[0122] In summary, in the embodiments of the present disclosure, after detecting a consultation question of the second object identifier for the first object, the interaction data of the first object identifier in the current session is obtained, where the interaction data is used to indicate the conversation content and the source of the conversation content between the second object identifier and the first object. Next, the historical task data executed by the second object identifier for the first object identifier can be obtained, and the data to be confirmed associated with the interaction data can be determined from the historical task data. Then, based on the data characteristics of the interaction data, the target task data associated with the consultation question can be predicted from the data to be confirmed, so as to determine the reply content corresponding to the consultation question according to the target task data, thereby predicting the target task data that hits the consultation question in the historical task data in combination with the interaction data, improving the accuracy of the prediction of the target task data, enabling the artificial intelligence assistant to provide correct help to the user, and further improving the user experience.

[0123] In this embodiment, an information prediction device is further provided. The device is used to implement the above embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" may be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0124] This embodiment provides an information prediction device, as Figure 5 shown, including:

[0125] An acquisition module 501, configured to obtain the interaction data of the first object identifier in the current session after detecting a consultation question of the second object identifier for the first object, where the interaction data is used to indicate the conversation content and the source of the conversation content between the second object identifier and the first object;

[0126] A determination module 502, configured to obtain the historical task data executed by the second object identifier for the first object identifier, and determine the data to be confirmed associated with the interaction data from the historical task data;

[0127] A prediction module 503, configured to predict the target task data associated with the consultation question from the data to be confirmed based on the data characteristics of the interaction data, so as to determine the reply content corresponding to the consultation question according to the target task data.

[0128] In some alternative implementation manners, the interaction operations corresponding to the interaction data include: the session triggering behavior of the second object identifier, and the conversation behavior between the second object identifier and the first object, where the session triggering behavior is used to indicate the behavior path for the second object identifier to trigger the session. The acquisition module 501 is further configured to:

[0129] Determine whether it is detected that the conversation behavior includes the sending behavior of the information entity;

[0130] If so, obtain the first historical task information and / or the first item information sent by the second object identifier to the first object during the session, where the interaction data includes the first historical task information and / or the first item information;

[0131] If not, obtain the behavior path data corresponding to the session trigger behavior, where the interaction data includes the behavior path data.

[0132] In some alternative embodiments, the interaction data includes behavior path data, and the obtaining module 501 is further configured to:

[0133] Based on the behavior path data, determine whether the second object identifier triggers a session behavior through historical task data;

[0134] If so, determine the historical task data triggered by the second object identifier as the data to be confirmed;

[0135] If not, when it is determined based on the behavior path data that the second object identifier triggers a session behavior through the item display entity, determine the data to be confirmed that matches the item display entity in the historical task data, where the item display entity is used to display the item information of the item that can be obtained.

[0136] In some alternative embodiments, the determining module 502 is further configured to:

[0137] When it is detected that the session behavior includes a sending behavior and the behavior path data is not obtained, determine the task data that meets the preset time condition in the historical task data as the data to be confirmed.

[0138] In some alternative embodiments, the interaction data includes: historical conversation content, interaction task information, where the historical conversation content includes at least one round of conversation content between the second object identifier and the first object, and the interaction task information is used to indicate the task information of the historical task data for the second object identifier to initiate a service request to the first object. The prediction module 503 is further configured to:

[0139] Extract the data features of the historical conversation content and the interaction task information;

[0140] Input the data features into the prediction model to obtain the target task data.

[0141] In some alternative embodiments, the prediction model includes a recommendation model and a language model. The prediction module 503 is further configured to:

[0142] Input the data features into the recommendation model to obtain the prediction results corresponding to each data to be confirmed;

[0143] When it is determined that the prediction result meets the adjustment condition, the language model is called to process the historical output result of the data to be confirmed, where the historical output result is used to indicate the prediction result of the historical task data discussed in the historical conversation content output by the language model;

[0144] Based on the historical output result, the prediction result is adjusted, and the target task data whose prediction result meets the matching degree condition is determined from the data to be confirmed.

[0145] In some alternative embodiments, the prediction module 503 is further configured to:

[0146] When the differences of the prediction results are all within a preset prediction interval, it is determined that the prediction result meets the adjustment condition.

[0147] In some alternative embodiments, the prediction module 503 is further configured to:

[0148] Increase a preset matching value for the prediction result corresponding to the data to be confirmed that matches the historical output result, and decrease the preset matching value for the prediction result corresponding to the data to be confirmed that does not match the historical output result.

[0149] The further functional descriptions of the above-mentioned various modules and units are the same as those in the corresponding embodiments above, and will not be elaborated here.

[0150] The information prediction device in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0151] This embodiment of the present disclosure further provides a computer device having the above Figure 5 shown information prediction device.

[0152] Please refer to Figure 6 , Figure 6 which is a schematic structural diagram of a computer device provided by an alternative embodiment of the present disclosure. As shown in Figure 6As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting the components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some alternative embodiments, if needed, multiple processors and / or multiple buses can be used together with multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (such as an array of servers, a set of blade servers, or a multi-processor system). Figure 6 In [the figure], a processor 10 is taken as an example.

[0153] The processor 10 can be a central processing unit, a network processor, or a combination thereof. Among them, the processor 10 can further include a hardware chip. The above-mentioned hardware chip can be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The above-mentioned programmable logic device can be a complex programmable logic device, a field-programmable gate array, a generic array logic, or any combination thereof.

[0154] Among them, the memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiments.

[0155] The memory 20 can include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the computer device, etc. In addition, the memory 20 can include high-speed random access memory, and can also include non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 can optionally include a memory remotely set relative to the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the above-mentioned network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0156] The memory 20 can include volatile memory, such as random access memory; the memory can also include non-volatile memory, such as flash memory, a hard disk, or a solid-state drive; the memory 20 can also include a combination of the above types of memory.

[0157] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0158] Embodiments of the present disclosure also provide a computer-readable storage medium. The methods according to the embodiments of the present disclosure can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code that is originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the methods described herein can be processed by such software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the methods shown in the above embodiments are implemented.

[0159] A part of the present invention can be applied as a computer program product, for example, computer program instructions, which when executed by a computer, can call or provide the methods and / or technical solutions according to the present invention through the operation of the computer. Those skilled in the art should be able to understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Herein, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to the computer.

[0160] Although the embodiments of the present disclosure are described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations all fall within the scope defined by the present disclosure.

Claims

1. An information prediction method, characterized in that: The method comprises: After detecting the consulting question of the second object identifier for the first object, acquiring the interaction data of the first object identifier in the current session, wherein the interaction data is used to indicate the content of the session and the source of the content of the session between the second object identifier and the first object; Acquire historical task data executed by the second object identifier for the first object identifier, and determine to-be-confirmed data associated with the interaction data in the historical task data; Based on the data features of the interaction data, the target task data associated with the consulting question in the data to be confirmed is predicted, so as to determine the reply content corresponding to the consulting question according to the target task data.

2. The method according to claim 1, characterized in that The interaction operation corresponding to the interaction data includes: a session triggering behavior of the second object identifier, and a session behavior between the second object identifier and the first object, wherein the session triggering behavior is used to indicate a behavior path of the second object identifier triggering a session; The acquiring the interaction data of the first object identifier in the current session includes: Determining whether it is detected that the session behavior includes a sending behavior of an information entity; If so, obtaining the first historical task data and / or the first item information sent by the second object identifier to the first object during the session; If not, obtain behavior path data corresponding to the session triggering behavior, wherein the interaction data includes the behavior path data.

3. The method according to claim 2, characterized in that The interaction data includes the behavior path data; The determining, in the historical task data, the data to be confirmed that is associated with the interaction data includes: Based on the behavior path data, determining whether the second object identifier triggers the session behavior through historical task data; If so, determining the historical task data triggered by the second object identifier as the data to be confirmed; If not, when it is determined based on the behavior path data that the second object identifier triggers the session behavior through an item display entity, the data to be confirmed that matches the item display entity is determined in the historical task data, wherein the item display entity is used to display item information of available items.

4. The method according to claim 2, characterized in that: The determining of the to-be-confirmed data associated with the interaction data in the historical task data further includes: When it is detected that the session behavior includes the sending behavior and the behavior path data is not obtained, the task data that meets the preset time condition in the historical task data is determined as the data to be confirmed.

5. The method according to claim 1, characterized in that The interaction data includes: historical conversation content and interaction task information, wherein the historical conversation content includes at least one round of conversation content between the second object identifier and the first object, and the interaction task information is used to indicate task information of a historical task in which the second object identifier initiates a service request to the first object; The predicting, based on the data features of the interaction data, the target task data associated with the consulting question in the to-be-confirmed data includes: Extracting data features of the historical conversation content and the interactive task information; The data features are input into the prediction model to obtain the target task data.

6. The method according to claim 5, characterized in that The prediction model includes a recommendation model and a language model; The step of inputting the data features into a prediction model to obtain the target task data includes: Input the data features into the recommendation model to obtain prediction results corresponding to each data to be confirmed; When it is determined that the prediction result satisfies the adjustment condition, calling the historical output result of the language model for the data to be confirmed, wherein the historical output result is used to indicate the prediction result output by the language model for the historical task data discussed in the historical conversation content; The prediction result is adjusted based on the historical output result, and target task data whose prediction result satisfies a matching condition is determined in the data to be confirmed.

7. The method according to claim 6, characterized in that The determining that the prediction result meets the adjustment condition includes: When the differences of the prediction results are all within a preset prediction interval, it is determined that the prediction results meet the adjustment condition.

8. The method according to claim 6, characterized in that The adjusting the prediction result based on the historical output result comprises: The predicted result corresponding to the data to be confirmed that matches the historical output result is increased by a preset matching value, and the predicted result corresponding to the data to be confirmed that does not match the historical output result is decreased by a preset matching value.

9. An information prediction device, characterized in that: The device comprises: An acquisition module, configured to acquire interaction data of the first object identifier in a current session after detecting a consulting question of the second object identifier for the first object, wherein the interaction data is used to indicate the content of the session and the source of the content of the session between the second object identifier and the first object; a determination module, configured to obtain historical task data executed by the second object identifier for the first object identifier, and determine to-be-confirmed data associated with the interaction data in the historical task data; A prediction module is used to predict the target task data associated with the consulting question in the data to be confirmed based on the data features of the interactive data, so as to determine the reply content corresponding to the consulting question according to the target task data.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the information prediction method according to any one of claims 1 to 8 by executing the computer instructions.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the information prediction method according to any one of claims 1 to 8.

12. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the information prediction method according to any one of claims 1 to 8.

Images